Blog Post < Previous
CVP
Paradigm Shift - Part VIII: Gen-AI Can Reduce Housing Costs
As discussed in Paradigm Shift - Part VII: Faster, Cheaper, Better, in many ways the technology being used by design profession is decades ahead of the housing construction business. Architects, engineers, designers, and professional construction managers are becoming fully integrated into the data-driven CAD and BIM software world. But in general, single-family and multifamily housing development has yet to be significantly disrupted by the advancements in Artificial Intelligence (AI) and Generative Artificial Intelligence (Gen-AI).
What is Generative AI?
Gen-AI is the output of software and microprocessor computational systems that can analyze vast amounts of information (“data”) and "generate" new content in different forms, based on the data it was trained on and the refinement and improvement of its reasoning capabilities derived from its interactions with its users, post training. (“machine learning”).
It is built on complex mathematical models and algorithms, primarily using artificial neural networks that mimic the neural structure of the human brain. It can recognize patterns and multi-dimensional relationships in enormous datasets that would take a human being years or even centuries to review. It can present responses in the form of text, images (ChatGPT, Perplexity, etc.), videos, software code, or other mediums.
And it can do it almost instantly.
Consider that Nvidia’s latest “Blackwell,” graphical processing unit (GPU), which is part of the core infrastructure of Gen-AI, can perform up to 20 quadrillion computations (Floating Point Operations) per second -- that’s one thousand trillion or one million, million, million computations per second per GPU -- and large AI data centers house thousands of these GPUs, working in parallel.
As such, Gen-AI can propose solutions that have never been imagined before. In this way, Gen-AI is different from traditional AI because it can create new things rather than just perform specific tasks or make predictions. (e.g., the odds of a ballplayer hitting a home run).
This ability to generate original content is what makes it so powerful. And it can be applied to anything, including solving affordable housing design, construction, and cost reduction challenges.
If Gen-Al is trained on or directed to search the internet to find everything there is to know about architecture, engineering, building codes and regulations (height limit, floor area rations, setbacks, etc.), construction methods, materials, and products, and then given a “program” for a project (i.e., multifamily affordable housing) and a location, it can use all the data available (e.g., weather, topography, hydrology, geology, solar cycles, property plat maps, etc. ) to iterate design concepts in response to requests and modifications and standards proposed by the architect/designer.
How this works
Let’s assume that a developer wants to build an affordable, infill, mixed-use, multifamily affordable housing project on a narrow, half-acre parcel adjacent to a creek. The developer’s concept is to create a mixed-use “Village” for artists and creative professionals. As such, the developer wants it to have a contemporary look and use readily available, low-maintenance materials and finishes.
But, the developer also wants the project to be made of prefabricated components that can be trucked to the building site. To the greatest extent possible, the developer wants those factory-built components to be “plug and play” and stitched together on-site and with minimal manual labor.
Communicating with a Gen-AI is called "prompting," an important skill to get good results. Working with a Gen-AI is an iterative process (trial and error) that proceeds in steps, to refine the output at each stage.
In a very simplified example, let's say that Gen-AI is asked to start by creating designs for bathroom and kitchen units of certain sizes with certain features and fixture types, and for a budgeted cost per square foot.
Click on the image to enlarge
Then, Gen-AI is asked to combine these into apartment plans of certain types and sizes: studios, one bedroom, and two bedroom units, using shared, modular dimensions and desired room square footage requirements.
Click on the image to enlarge
Finally, we ask Gen-AI to combine all of these components into a design for the mixed-use “Village” concept, having the maximum number of units allowable within the site’s physical limitations and the regulatory restrictions in the local codes.
However, we also want the building to include a "dramatic" interior “street” filled with natural light, a small café/coffee bar, some storefronts with live/work apartments above, an art gallery that is open to the public, a shared studio space for residents, storage units, and other features… and we want it all to have a “residential” scale and for the roof lines to be “interesting"… and to have solar panels on the roof to power the public space energy needs.
After many iterations, the result might look something like this.
Street View
First floor plan
Second floor plan
3-D model
The design concept can then be further refined and manipulated by prompting Gen-AI to prioritize different design criteria, resulting in adjustments to the building’s form and materials to reflect the prompts.
For instance, a designer might ask Gen-AI to,
“Optimize the building form and materials to maximize passive solar gain and lower heating and cooling energy costs,” or
“Which type of structural system (concrete, steel, wood frame, etc.) will reduce the lifecycle, carbon emissions of the building materials used” and enhance the building’s overall environmental sustainability? or
“Adjust the floor plans to be the most efficient use of space within the allowable building envelope.”
“Now, combine all of these requirements equally into the project.”
And since this design data all resides in “Building Information Management” (BIM) software (in this instance, the model was built in Autodesk Revit), the refinements and iterations can keep going down to the smallest detail, even the interior design and furnishings used in model units to market the property to prospective renters or buyers.
This also allows us to view images of the final design.
Creekside view
Views of interior/exterior public spaces
Views of apartment interiors
However, the scope of the challenges Gen-AI can address goes even further than this.
For example, once the project design is finalized Gen-AI could be asked to determine how to fast-track the construction management workflow and scheduling between all the project participants (architects, engineers, contractors, material suppliers, component fabricators, etc.), from the design concept phase to the final move-in date.
The logical extension of all this is to maximize efficiencies and minimize costs at every step is by extending Gen-AI project workflow optimization solutions into “factory-built” production lines and by implementing Gen-AI-driven robotics.
This start-to-finish integration can unlock the enormous potential of Gen-AI to maximize the designer’s ability to create unique, customizable, appropriately scaled, affordable housing projects at the lowest possible cost without sacrificing amenities or quality.
I understand that this may seem like science fiction to anyone who’s developed multifamily housing, but these concepts are already being successfully implemented in other industries.
The challenge is how to make it a reality.
NEXT – Gen-AI Factory Built
For more, see;
Paradigm Shift: Rethinking Housing Affordability
Paradigm Shift - Part II: Housing Unaffordability May Be Just Beginning
Paradigm Shift - Part III: How Affordable Housing Need Powered the Modernist Movement
Paradigm Shift – Part IV: The Assault on the American Dream
Paradigm Shift - Part V: Automation and AI, Double-edged Swords for the Housing Industry
Paradigm Shift - Part VI: New Hope for Affordable Housing?
Paradigm Shift - Part VII: Faster, Cheaper, Better
Bob Silvestri is a long-time Mill Valley resident, the Editor of the Marin Post, and the founder and president of Community Venture Partners, a 501(c)(3) nonprofit community organization dedicated to bringing the community's voice to government decision-making and funded exclusively by individuals and nonprofit donors. Please consider donating now to support our work.